Did early login help in NSE TBT architecture?
Did the early bird get the worm? Did the brokers who logged in first make more money? It is surprising that after 4 years of investigation, this question is still being asked. Various experts have understood the NSE system and have based their findings on understanding, forensic examination of code and even simulations. In this article we try to find if there was any direct evidence of first login advantage.
This article will use some technical words which I will define below:
Tick: An event of the market which is sent to the members as data feed. This was sent as TCP packet in TCP TBT times and is now sent as UDP packets in MTBT. Ticks are sent in sequence and each tick has a unique tick sequence number. This sequence number is generated by Exchange and therefore is recorded in exchange data.
Trigger Tick: Not all ticks trigger order after the Member algorithm has done its processing. A tick which triggers a unique order from a Member’s algorithm is called a Trigger Tick. So, for every order there is a unique trigger tick with a unique tick sequence number. As all software is written keeping back testing in mind, the trigger ticks are recorded in the Member’s database.
Common Orders: Multiple orders from multiple servers triggered by a common tick. If a single tick triggers orders across multiple trading servers of one or more members, then all such orders can be referred to as common orders.
Exchange Order Number: Each order when it reaches the exchange gets assigned a unique number called exchange order number. The order which reaches the exchange matching engine earlier in time has lower exchange order number.
Earlier, when news of the NSE colocation “scam” had started pouring in, I didn’t know much about NSE architecture and my first impression also was that a huge problem had been detected at NSE. However, the hard data has presented a totally different picture.
A tick in NSE goes through a complex system involving multiple layers of hardware and software. These systems are common on exchange side but can be very different on the member side. If we consider this entire system as a black box, the entire premise of the allegations in the NSE collocation “scam” is that first to login is first to trade. This can be further simplified to, first to login has the lowest exchange order number for common orders.
The biggest issue while looking at the direct evidence is that different members use different systems. So how can we say whether the lower exchange order number for common orders is due to differentiation brought in by members’ systems or due to first login? The only way to eliminate this bias is to compare the common orders of identical trading servers running identical algorithms. Therefore, the only way to carry out this analysis is to compare trading servers of the same trading member for common orders/opportunities.
One of the SEBI noticees in the NSE Colo matter has done exactly this. They have carried out absolutely relevant data analysis to demonstrate if the servers which logged in a particular time sequence, also get exchange order numbers for common orders in the same time sequence. If this was found to be true, this would be clinching evidence to prove that there indeed was benefit as alleged by the whistle-blower. However their results show exactly the opposite. It has clearly been demonstrated on the basis of hard data by this noticee that there was no correlation between order numbers of common orders fired by different servers and their respective login time. This clearly demonstrates by direct evidence that there was no benefit because of first or early login. This is not surprising at all for reasons which have been explained in detail in my earlier blogs in this series.
This noticee has not only submitted detailed data analysis on these lines, they have also submitted expert opinion given by a renowned Professor of IIT. This expert opinion submitted by the noticee goes on to provide theoretical reasoning for this finding of the analysis of data. The opinion not only provides empirical evidence of the behavior but also simulation under various conditions. The noticee claims that the inherent variability in the NSE systems was so large that there was no possibility of any advantage by logging in first or early. Similar data analysis has been conducted for logging in to secondary versus primary and it has been demonstrated by hard data based direct evidence that there was no advantage in logging in to secondary.
With such direct evidence being available, it is surprising that none of the SEBI appointed investigators looked into it. It is very easy for Deloitte or E&Y to validate if the findings of this expert and this noticee are indeed correct or not. The Government has spent humongous amount of time and money on this investigation across multiple agencies. If this direct evidence really points to NO advantage, would the Government try to question the real people behind making a mountain of a molehill in the NSE colocation case; probably to deflect attention from the real scam that had happened just in the earlier year and in which thousands of crores was lost by thousands of investors? This data analysis does not of course address all the allegations against NSE. There are many more allegations against NSE.
All articles are here. The author advises market participants in legal matters related to securities markets and has advised some noticees in this matter also.
